A guide to computerized Trade Management

This is more of a theoretical question re correlation of systems. When building systems into a portfolio, it seems to me that I would want them to all have a correlation in the sense that I want a straight equity curve and that one systems drawdown is canceled by another systems run up. With this in mind, it would seem to be better to correlate the underwater equity curves and look for uncorrelated systems rather than correlating the equity curves. I have tried both methods using Excel and I found that the uncorrelated drawdown of systems gave a smoother equity curve for the portfolio.
 
Quote from rdencpa:

This is more of a theoretical question re correlation of systems. When building systems into a portfolio, it seems to me that I would want them to all have a correlation in the sense that I want a straight equity curve and that one systems drawdown is canceled by another systems run up. With this in mind, it would seem to be better to correlate the underwater equity curves and look for uncorrelated systems rather than correlating the equity curves. I have tried both methods using Excel and I found that the uncorrelated drawdown of systems gave a smoother equity curve for the portfolio.
More...

That interesting idea which I have not looked at. I will look at it and see if I want to add it to TradersStudio.
 
Quote from Murray Ruggiero:

What other examples would you like to see ?
More...
Do you use Jurik's WAV data preprocessor in your system, and can you explain what it outputs? Thanks in advance.
 
Quote from kut2k2:

Do you use Jurik's WAV data preprocessor in your system, and can you explain what it outputs? Thanks in advance.
More...

I don't use Jurik Wav preprocessing in my systems. I do use some concepts of it. They were published in the early 1990's in Futures magazine. Mark has advanced this preprocessing over the years since then.

Wav saves a lot of time in developing Neural network by automating input development and for 90% of the people this type of canned preprocssing is a good way to get some good results quickly. In fact I would often use it to get quick and dirty results to test a new target before developing my own custom preprocessing. I have used Wav as a baseline.

If you have a lot of domain expertise and expertise working with neural network , I feel you can develop a better model using custom inputs. The problem is only a small percentage of the population can do this.
 
Quote from Murray Ruggiero:
If you are registered on the web site you can go onto the document tab and download the language reference on line help.
More...
Does the online help refer to the latest version available?
 
Quote from JaiSreeram:

Does the online help refer to the latest version available?
More...

The online help you can download is for 2.5.5 which is the current shipping version. We have a new help for the beta version which is not ready yet. We will not update the help available on the site until that version is released.
 
You must log in or register to reply here.
Back
Top